Config 'default' not found for dataset

[hakatashi]

While adding a dataset hakatashi/hakatashi-pixiv-bookmark-deepdanbooru to my project, I encountered the following error.

Config 'default' not found for dataset 'hakatashi/hakatashi-pixiv-bookmark-deepdanbooru', which has the following configs: ['hakatashi--hakatashi-pixiv-bookmark-deepdanbooru']

But the dataset definitely has default config. README.md · hakatashi/hakatashi-pixiv-bookmark-deepdanbooru at main

---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: validation
    path: data/validation-*
dataset_info:
  features:
  - name: tag_probs
    sequence: float32
  - name: class
    dtype:
      class_label:
        names:
          '0': not_bookmarked
          '1': bookmarked_public
          '2': bookmarked_private
  splits:
  - name: train
    num_bytes: 4301053452
    num_examples: 179121
  - name: test
    num_bytes: 1433684484
    num_examples: 59707
  - name: validation
    num_bytes: 1433708496
    num_examples: 59708
  download_size: 7351682183
  dataset_size: 7168446432
task_categories:
- image-classification
- tabular-classification
tags:
- art
size_categories:
- 100K<n<1M
---

What went wrong? I feel like this is a bug.

[claytonsmith]

Where you able to figure it out?

[sugreev]

Having exactly the same problem! Autotrain wasn’t even able to create my dataset automatically (see post here), so I tried making one manually. But now Autotrain won’t correctly load it even though I included a Config YAML block in my readme file?? Seems faster to just manually train with all these Autotrain bugs, especially with such limited docs

[abhishek]

You should use the space for training. The UI is deprecating soon and wont be updated. docs are here: AutoTrain. Use the spaces method.